Nomination and inaugural speeches are the first step of Presidents. These speeches indicate what Presidents advocate and what’s intended policy they would conduct during tenure, which undoubtedly affect the decision of voters. However, some Presidents owned only one term while others perfromed two-or-more terms? What’s the reason? Is it possible that these can be seen from their nominaiton and inaugural speeches, which convey their thoughts directly to voters? In this project, we would analyze the nomination and inaugural speeches of past presidents, and try to figure out whether there is any differences between those one-term Presidens and multi-terms Presidents in their speeches, or on the opposite way, their differences cannot be figured out from just speeches.

Step I: Environment preparation — check and install needed packages.

packages.used=c("rvest", "tibble", "qdap", "sentimentr", 
                "gplots", "dplyr","tm", "syuzhet", 
                "factoextra", "gridExtra", "scales", "RColorBrewer",
                "RANN", "topicmodels","wordcloud","tidytext","ggridges","ggplot2")

# check packages that need to be installed.
packages.needed=setdiff(packages.used, 
                        intersect(installed.packages()[,1], packages.used))

# install additional packages
if(length(packages.needed)>0){
  install.packages(packages.needed, dependencies = TRUE,
                   repos='http://cran.us.r-project.org')
}

# load packages
library("rvest")
library("tibble")
library("qdap")
library("sentimentr")
library("gplots")
library("dplyr")
library("tm")
library("syuzhet")
library("factoextra")
library("scales")
library("RColorBrewer")
library("RANN")
library("topicmodels")
library("wordcloud")
library("tidytext")
library("gridExtra")
library('ggridges')
library('ggplot2')

source("../lib/plotstacked.R")
source("../lib/speechFuncs.R")

Step II: Data collection and first-step processing.

In this project, we select all inaugural and some of nomination speeches of past presidents.

Scrape speech URLs from http://www.presidency.ucsb.edu/
## Inaugural speeches
# Get link URLs
main.page <- read_html(x = "http://www.presidency.ucsb.edu/inaugurals.php")
inaug=f.speechlinks(main.page)
tail(inaug,5)
##                          links
## 55            January 20, 2005
## 56            January 20, 2009
## 57            January 21, 2013
## 58            January 20, 2017
## 59 click to see a larger image
##                                                      urls
## 55  http://www.presidency.ucsb.edu/ws/index.php?pid=58745
## 56     http://www.presidency.ucsb.edu/ws/index.php?pid=44
## 57 http://www.presidency.ucsb.edu/ws/index.php?pid=102827
## 58 http://www.presidency.ucsb.edu/ws/index.php?pid=120000
## 59                      images/inaugurals-words-large.jpg
inaug=inaug[-nrow(inaug),]   # Remove the last line, irrelevant due to error

## Nomination speeches
main.page=read_html("http://www.presidency.ucsb.edu/nomination.php")
# Get link URLs
nomin <- f.speechlinks(main.page)
tail(nomin,5)
##                 links
## 51   William McKinley
## 52  Benjamin Harrison
## 53  Benjamin Harrison
## 54 James A. Garfield 
## 55   Abraham Lincoln 
##                                                     urls
## 51 http://www.presidency.ucsb.edu/ws/index.php?pid=76197
## 52 http://www.presidency.ucsb.edu/ws/index.php?pid=76067
## 53 http://www.presidency.ucsb.edu/ws/index.php?pid=76068
## 54 http://www.presidency.ucsb.edu/ws/index.php?pid=76221
## 55 http://www.presidency.ucsb.edu/ws/index.php?pid=88863
Prepare CSV datasets for the scraped speeches from speech metadata posted on http://www.presidency.ucsb.edu/
inaug.list=read.csv("../data/inauglist.csv", stringsAsFactors = FALSE)
nomin.list=read.csv("../data/nominlist.csv", stringsAsFactors = FALSE)

speech.list=rbind(inaug.list, nomin.list)
speech.list$type=c(rep("inaug", nrow(inaug.list)),
                   rep("nomin", nrow(nomin.list)))
nomin<-nomin[-47,]   #Delete a redundant row in nomin
speech.url=rbind(inaug, nomin)
speech.list=cbind(speech.list, speech.url)  #Combine original list with URLs
Scrap the main texts of speeches from the speech URLs
speech.list$fulltext=NA
for(i in 1:nrow(speech.list)) {
    text <- read_html(speech.list$urls[i]) %>%   #Load the page
    html_nodes(".displaytext") %>%    #Isloate the text
    html_text()    #Get the text
    speech.list$fulltext[i]=text
    # Create the file name
    filename <- paste0("../data/fulltext/", 
                     speech.list$type[i],
                     speech.list$File[i], "-", 
                     speech.list$Term[i], ".txt")
    sink(file = filename) %>%    #Open file to write 
    cat(text)    #Write the file
    sink()    #Close the file
}
Subset the dataset

In order to dicover any differences between one-term Presidents and multi-terms Presidents in their inaugural and nomination speeches, We have first to subset the whold speech dataset into several small directed parts, such as prepare a dataset for inaugural speeches of those one-term presidents as single_term_inaug_speech.list.

## Inaugural speeches
first_inaug_speech.list<-speech.list %>% filter(type=='inaug',Term==1)
first_inaug_speech.list[18,'President']<-'Grover Cleveland'
first_inaug_speech.list[18,'File']<-'GroverCleveland'

secondormore_inaug_speech.list<-speech.list %>% filter(type=='inaug',Term>=2)
secondormore_inaug_speech.list[8,'President']<-'Grover Cleveland'
secondormore_inaug_speech.list[8,'File']<-'GroverCleveland'

multi_terms_inaug_files<-unique(secondormore_inaug_speech.list[,'File'])
# The first inaugural speech for the Presidents with multi-terms
multi_terms_inaug_speech1.list<-first_inaug_speech.list %>% filter(File%in%multi_terms_inaug_files)
single_term_inaug_speech.list<-first_inaug_speech.list %>% filter(!(File%in%multi_terms_inaug_files))
  
## Nomination speeches
single_term_presid<-unique(single_term_inaug_speech.list[,'President'])
multi_terms_presid<-unique(multi_terms_inaug_speech1.list[,'President'])

first_nomin_speech.list<-speech.list %>% filter(type=='nomin',Term==1)
secondormore_nomin_speech.list<-speech.list %>% filter(type=='nomin',Term>=2, President%in%multi_terms_presid)
# The first nomination speech for the Presidents with multi-terms
multi_terms_nomin_speech1.list<-first_nomin_speech.list %>% filter(President%in%multi_terms_presid)
single_term_nomin_speech.list<-first_nomin_speech.list %>% filter(President%in%single_term_presid)

Step III: Word Frequency Analysis

What would be the differnces between one-term Prsidents and multi-terms Presidents in their inaugural and nomination speeches? Let’s start from the simplest aspect. What about the word frequency? Is it possible that those one-term Presidents prefered to use some kinds of wrods, while those multi-terms Presients tend to have an opposites way? Let’s find something interesting.

First of all, for speeches we first need to clean them up, such as remove those whitespace, change all the letters into lower cases, and remove english common stopwords, etc.

Clean the speeches:
## Create a function to general clean those text datasets
clean_dataset<-function(list, after){
  after<-Corpus(VectorSource(list))
  after<-tm_map(after, stripWhitespace)
  after<-tm_map(after, content_transformer(tolower))
  after<-tm_map(after, removeNumbers)
  after<-tm_map(after, removeWords, stopwords('english'))
  after<-tm_map(after, removeWords, character(0))
  after<-tm_map(after, removePunctuation)
  after<-tm_map(after, stemDocument)
  return(after)
}

## First inaugural speeches (single-term)
single_inaug<-clean_dataset(single_term_inaug_speech.list$fulltext, single_inaug)

## First inaugural speeches (multi-terms)
multi_first_inaug<-clean_dataset(multi_terms_inaug_speech1.list$fulltext, multi_first_inaug)

## Second inaugural speeches (multi-terms)
multi_second_inaug<-clean_dataset(secondormore_inaug_speech.list$fulltext, multi_second_inaug)

## First nomination speeches (single-term)
single_nomin<-clean_dataset(single_term_nomin_speech.list$fulltext, single_nomin)

## First nomination speeches (multi-term)
multi_first_nomin<-clean_dataset(multi_terms_nomin_speech1.list$fulltext, multi_first_nomin)

## Second nomination speeches (second-term)
multi_second_nomin<-clean_dataset(secondormore_nomin_speech.list$fulltext, multi_second_nomin)
Build a term-document matrix:

Document matrix is a table containing the frequency of the words.

## Create a function to build term-document matix and return a dataframe with frequency
build_tdm<-function(corpus, df_name){
  tdm<-TermDocumentMatrix(corpus)
  m<-as.matrix(tdm)
  v<-sort(rowSums(m), decreasing=TRUE)
  df_name<-data.frame(word=names(v), freq=v, row.names=1:length(v))
  return(df_name)
}

## Inaugural speeches
single_inaug_tdm<-build_tdm(single_inaug, single_inaug_tdm)
head(single_inaug_tdm, 10)
##         word freq
## 1       will  426
## 2     govern  342
## 3     nation  328
## 4      peopl  298
## 5      state  259
## 6        can  229
## 7       upon  215
## 8      power  214
## 9    countri  210
## 10 constitut  185
multi_first_inaug_tdm<-build_tdm(multi_first_inaug, multi_first_inaug_tdm)
head(multi_first_inaug_tdm, 10)
##      word freq
## 1    will  276
## 2  govern  169
## 3  nation  164
## 4   peopl  161
## 5     can  143
## 6    must  105
## 7   state  103
## 8   great   97
## 9   shall   94
## 10  world   93
multi_second_inaug_tdm<-build_tdm(multi_second_inaug, multi_second_inaug_tdm)
head(multi_second_inaug_tdm, 10)
##      word freq
## 1    will  230
## 2  nation  183
## 3   peopl  159
## 4  govern  144
## 5   world  100
## 6     can   97
## 7   great   93
## 8   state   93
## 9   power   92
## 10   time   91
## Nomination speeches
single_nomin_tdm<-build_tdm(single_nomin, single_nomin_tdm)
head(single_nomin_tdm, 10)
##        word freq
## 1      will  302
## 2    nation  188
## 3     peopl  188
## 4  american  163
## 5    govern  163
## 6     parti  138
## 7   countri  133
## 8       law  129
## 9       can  128
## 10      one  125
multi_first_nomin_tdm<-build_tdm(multi_first_nomin, multi_first_nomin_tdm)
head(multi_first_inaug_tdm, 10)
##      word freq
## 1    will  276
## 2  govern  169
## 3  nation  164
## 4   peopl  161
## 5     can  143
## 6    must  105
## 7   state  103
## 8   great   97
## 9   shall   94
## 10  world   93
multi_second_nomin_tdm<-build_tdm(multi_second_nomin, multi_second_nomin_tdm)
head(multi_second_inaug_tdm, 10)
##      word freq
## 1    will  230
## 2  nation  183
## 3   peopl  159
## 4  govern  144
## 5   world  100
## 6     can   97
## 7   great   93
## 8   state   93
## 9   power   92
## 10   time   91

Let’s see more clear in Word Cloud plots.

Generate the Word Cloud:
par(oma=c(1,1,2,1),bg='beige')
layout(rbind(1,cbind(2,3)))
## Inaugural speeches
set.seed(2017)
wordcloud(single_inaug_tdm$word, freq=single_inaug_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Single-term',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

wordcloud(multi_first_inaug_tdm$word, freq=multi_first_inaug_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Multi-terms(1st)',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

wordcloud(multi_second_inaug_tdm$word, freq=multi_second_inaug_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Multi-terms(2nd)',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

mtext('Wordclouds of Inaugural Speeches of Single and Multi-terms',
      side=3,line=-2,cex=1.8,font=2,outer=TRUE,col='navyblue')

par(oma=c(1,1,2,1),bg='beige')
layout(rbind(1,cbind(2,3)))
##Nomination speeches
wordcloud(single_nomin_tdm$word, freq=single_nomin_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Single-term',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

wordcloud(multi_first_nomin_tdm$word, freq=multi_first_nomin_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Multi-terms(1st)',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

wordcloud(multi_second_nomin_tdm$word, freq=multi_second_nomin_tdm$freq, max.words=100, min.freq=3,
          random.order=FALSE, rot.per=0.35, colors=brewer.pal(8,'Dark2'), use.r.layout=TRUE)
mtext('Multi-terms(2nd)',side=1,at=0.5,line=2,cex=1,font=2,col='navyblue')

mtext('Wordclouds of Nomination Speeches of Single and Multi-terms',
      side=3,line=-2,cex=1.8,font=2,outer=TRUE,col='navyblue')

As we can see, there is no much distinct difference in inaugural speeches, they all indicated much about anticipations and concerns about the development of whole nation and people, also the responsibility of governement. However, except those, Presidents who had only one term were more likely to emphasize something about ‘power’ in their inaugural speeches, which is less common in those multi-terms’ inaugural speeches.

In nomination speeches, somthing chages a little. Presidents who had multi-terms prefered to metion the word ‘american’, which perhaps indicated more ssense of unity than just ‘people’, more ofter than in those one-term’s speeches. Additionally, it is obvious that in multi-terms’ speeches, Presidents refered much to the word ‘new’, which appeared more a less in those one-term speeches. One more thing is that, in multi-terms’ speeches, since those Presidents had finished their first term, they always mentioned the word, ‘president’, in their speeches, that is because they hoped to remind people their successful President-time before, which can increase their confidences.

Step IV: Analysis of length of sentences

What kinds of sentences do President prefer? Short sentence is more likely to motivate people’s excetiments and morales, while long sentence tends to be more reliable and unstandable. Would those one-term Presidents tend to use longer sentence with many words than those Presidents owned muti-terms? Let’s find somthing interesting from this aspect.

Generate the lists of sentences
## Build a function that can create sentence lists and get emotions and valence
create_sent_list<-function(list,sent_list){
  for(i in 1:nrow(list)){
  sentences=sent_detect(list$fulltext[i],
                        endmarks=c('?', '.', '!', '|', ';'))
  if(length(sentences)>0){
    # get emotions and valence from NRC dictionary
    emotions=get_nrc_sentiment(sentences)
    word.count=word_count(sentences)
    emotions=diag(1/(word.count+0.01))%*%as.matrix(emotions)
    sent_list=rbind(sent_list, 
                    cbind(list[i,-ncol(list)], 
                          sentences=as.character(sentences), 
                          word.count,
                          emotions,
                          sent.id=1:length(sentences)))
  }
  }
  return(sent_list)
}
## Inaugural speeches
single_term_inaug_sent.list<-NULL
single_term_inaug_sent.list<-create_sent_list(single_term_inaug_speech.list,
                                              single_term_inaug_sent.list)%>%filter(!is.na(word.count)) 
multi_terms_inaug_sent1.list<-NULL
multi_terms_inaug_sent1.list<-create_sent_list(multi_terms_inaug_speech1.list,
                                               multi_terms_inaug_sent1.list)%>%filter(!is.na(word.count))
multi_terms_inaug_sent2.list<-NULL
multi_terms_inaug_sent2.list<-create_sent_list(secondormore_inaug_speech.list,
                                               multi_terms_inaug_sent2.list)%>%filter(!is.na(word.count))
## Nomination speeches
single_term_nomin_sent.list<-NULL
single_term_nomin_sent.list<-create_sent_list(single_term_nomin_speech.list,single_term_nomin_sent.list)%>%filter(!is.na(word.count))%>%filter(!is.na(File))

multi_terms_nomin_sent1.list<-NULL
multi_terms_nomin_sent1.list<-create_sent_list(multi_terms_nomin_speech1.list,
                                               multi_terms_nomin_sent1.list)%>%filter(!is.na(word.count))
multi_terms_nomin_sent2.list<-NULL
multi_terms_nomin_sent2.list<-create_sent_list(secondormore_nomin_speech.list,
                                          multi_terms_nomin_sent2.list)%>%filter(!is.na(word.count))
Create ridge line plots for the number of words in each sentence:
## Inaugural speeches
ggplot(data=single_term_inaug_sent.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Inaugural Speech (single-term)')
## Picking joint bandwidth of 5

ggplot(data=multi_terms_inaug_sent1.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Inaugural Speech (multi-terms (1st))')
## Picking joint bandwidth of 5.39

ggplot(data=multi_terms_inaug_sent2.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Inaugural Speech (multi-terms (2nd))')
## Picking joint bandwidth of 5.35

From ridge plots shown above, we can see that, in inaugural speeches, Presidents who owned only one term averagely tended to have 15 words in each sentence, and most of the sectences focused on 10-25 words, except some Presidents, like James Madison and Zachary Taylor, who also prefered longer sentences with more than 30 words per sentence, as well as completely diverse length of sentence in whold speeches. While for those presidents who owned multi-terms, the average length of sentence in their speeches shifts towards right a little bit, which is roughly 20 words per sentence, and what’s more, most of them prefered diverse length of sentence rather than all short ones or long ones.

par(mfrow=c(1,3))
## Nomination speeches
ggplot(data=single_term_nomin_sent.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Nomination Speech (single-term)')
## Picking joint bandwidth of 3.6

ggplot(data=multi_terms_nomin_sent1.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Nomination Speech (multi-terms (1st))')
## Picking joint bandwidth of 4.38

ggplot(data=multi_terms_nomin_sent2.list, aes(x=word.count, 
                                             y=reorder(President,word.count,mean)))+
  geom_density_ridges(fill='khaki1',alpha=0.5)+
  theme(axis.text.y=element_text(size=9), axis.text.x=element_text(face='bold',size=9),
        plot.title=element_text(face='bold.italic',size=15))+
  scale_x_continuous(breaks=c(seq(0,50,10),seq(50,125,25)),labels=c(seq(0,50,10),seq(50,125,25)))+
  labs(x='Numbe of Words / Per Sentence',y='',title='Nomination Speech (multi-terms (2nd))')
## Picking joint bandwidth of 3.4

From ridge plots shown above, we can see that, in one-term nomination speeches, Presidents prefered to have roughly 10 words in each sentence, and most of the sectences focused on 5-20 words. While in the first-term speech for president who owned multi-terms, the average length of sentence in their speeches shifts towards right a little bit, which is roughly 15 words per sentence, and the diversity of the length of sentences compacted a lot. Additionally, in the second-or-more-term speeches, those Presidents prefered more less words than before, the average number of words per sentence shift towards left a little.

Step V: Sentiment Analysis

We have evaluate the differences in word frequencies and length of sentences. What about the sentiments inside the speeches? How the Presidents shift between different setiments in their speeches. Is their any differnce among their terms?

Examples for sentiments charged in sentences
## Inaugural speeches
print("Franklin D. Roosevelt")
## [1] "Franklin D. Roosevelt"
speech.df=tbl_df(multi_terms_inaug_sent1.list)%>%
  filter(File=="FranklinDRoosevelt", word.count>=5)%>%
  select(sentences, anger:trust)
speech.df=as.data.frame(speech.df)
as.character(speech.df$sentences[apply(speech.df[,-1], 2, which.max)])
## [1] "Happiness lies not in the mere possession of money;"                 
## [2] "Happiness lies not in the mere possession of money;"                 
## [3] "Yet our distress comes from no failure of substance."                
## [4] "These are the lines of attack."                                      
## [5] "it lies in the joy of achievement, in the thrill of creative effort."
## [6] "We are stricken by no plague of locusts."                            
## [7] "Yet our distress comes from no failure of substance."                
## [8] "In this dedication of a Nation we humbly ask the blessing of God."
## Nomination speeches
print("William J. Clinton")
## [1] "William J. Clinton"
speech.df=tbl_df(multi_terms_inaug_sent2.list)%>%
  filter(File=="WilliamJClinton", word.count>=5)%>%
  select(sentences, anger:trust)
speech.df=as.data.frame(speech.df)
as.character(speech.df$sentences[apply(speech.df[,-1], 2, which.max)])
## [1] "We will stand mighty for peace and freedom and maintain a strong defense against terror and destruction."                                                    
## [2] "Fellow citizens, we must not waste the precious gift of this time."                                                                                          
## [3] "The divide of race has been America's constant curse."                                                                                                       
## [4] "They fuel the fanaticism of terror."                                                                                                                         
## [5] "May God strengthen our hands for the good work ahead, and always, always bless our America."                                                                 
## [6] "It was extended and preserved in the 19th century, when our Nation spread across the continent, saved the Union, and abolished the awful scourge of slavery."
## [7] "Fellow citizens, we must not waste the precious gift of this time."                                                                                          
## [8] "Let us meet them with faith and courage, with patience and a grateful, happy heart."
par(mfrow=c(1,2),mar=c(2,4,4,2))
# Inaugural speeches
single_term_inaug_emo.mean<-colMeans(select(single_term_inaug_sent.list, anger:trust)>0.01)
multi_terms_inaug_emo.mean1<-colMeans(select(multi_terms_inaug_sent1.list, anger:trust)>0.01)
multi_terms_inaug_emo.mean2<-colMeans(select(multi_terms_inaug_sent2.list, anger:trust)>0.01)
inaug_emo.mean<-as.data.frame(cbind(single_term_inaug_emo.mean, multi_terms_inaug_emo.mean1, 
                      multi_terms_inaug_emo.mean2),
                      col.names=c('single-term','multi-terms(1st)','multi-terms(2nd)'))

Col<-c('coral3','goldenrod1','darkolivegreen3')
barplot(t(inaug_emo.mean),col=Col,beside=TRUE,ylim=c(0,0.7),border=FALSE,las=1)
segments(1,seq(0,0.7,0.1),251,seq(0,0.7,0.1),lwd=1,col='lightgrey')
legend(x=3,y=0.7,fill=Col,bty='n',cex=1, 
       legend=c('single-term','multi-terms(1st)','multi-terms(2nd)'))
mtext('Inaugural Speeches',side=3,line=1,cex=1.5,font=2)

#Nomination speeches
single_term_nomin_emo.mean<-colMeans(select(single_term_nomin_sent.list, anger:trust)>0.01)
multi_terms_nomin_emo.mean1<-colMeans(select(multi_terms_nomin_sent1.list, anger:trust)>0.01)
multi_terms_nomin_emo.mean2<-colMeans(select(multi_terms_nomin_sent2.list, anger:trust)>0.01)
nomin_emo.mean<-as.data.frame(cbind(single_term_nomin_emo.mean, multi_terms_nomin_emo.mean1, 
                      multi_terms_nomin_emo.mean2),
                      col.names=c('single-term','multi-terms(1st)','multi-terms(2nd)'))

Col<-c('coral3','goldenrod1','darkolivegreen3')
barplot(t(nomin_emo.mean),col=Col,beside=TRUE,ylim=c(0,0.7),border=FALSE,las=1)
segments(1,seq(0,0.7,0.1),251,seq(0,0.7,0.1),lwd=1,col='lightgrey')
legend(x=3,y=0.7,fill=Col,bty='n',cex=1, 
       legend=c('single-term','multi-terms(1st)','multi-terms(2nd)'))
mtext('Nomination Speeches',side=3,line=1,cex=1.5,font=2)

From the grouped barplot shown above, we can easily see that in both inaugural and nomination speeches, Presidents who owned only one term more obviously displayed their sentiments in speeches than those with multi-terms, especially anger and trust in both kinds of speeches, and disgust, fear and sadness in nomination speeches. This outcome is more than interesting and make us think about that whether totally display those sentiments towards voters would affect the continuous-term of Presidents? Perhaps more revealable sentiments would distort the character of Presidents towards voters.

Whether those subtle differences of sentiment can roughly classify all the Presidents into two groups? We use the mean values of 8 different sentiments to do k-means cluster analysis, especially set k = 2.

# Inaugural speeches
single_term_inaug_sent.list$term_type<-'single'
multi_terms_inaug_sent1.list$term_type<-'multi'
multi_terms_inaug_sent2.list$term_type<-'multi'
inaug_sent.list<-as.data.frame(rbind(single_term_inaug_sent.list, multi_terms_inaug_sent1.list,
                                     multi_terms_inaug_sent2.list))
inaug_presid.summary=tbl_df(inaug_sent.list)%>%
  group_by(File)%>%
  summarise(
    anger=mean(anger),
    anticipation=mean(anticipation),
    disgust=mean(disgust),
    fear=mean(fear),
    joy=mean(joy),
    sadness=mean(sadness),
    surprise=mean(surprise),
    trust=mean(trust),
    negative=mean(negative),
    positive=mean(positive),
    term_type=unique(term_type)
  )
inaug_presid.summary=as.data.frame(inaug_presid.summary)
rownames(inaug_presid.summary)=as.character(inaug_presid.summary[,1])
km.res=kmeans(inaug_presid.summary[,-c(1,ncol(inaug_presid.summary))], iter.max=200, centers=2)
fviz_cluster(km.res, stand=FALSE, repel= TRUE,
             data = inaug_presid.summary[,-c(1,ncol(inaug_presid.summary))], 
             xlab='', xaxt='n', ylab='', show.clust.cent=FALSE)

table(inaug_presid.summary$term_type,as.vector(km.res$cluster))
##         
##           1  2
##   multi   7 10
##   single 10 12
# Nomination speeches
single_term_nomin_sent.list$term_type<-'single'
multi_terms_nomin_sent1.list$term_type<-'multi'
multi_terms_nomin_sent2.list$term_type<-'multi'
nomin_sent.list<-as.data.frame(rbind(single_term_nomin_sent.list, multi_terms_nomin_sent1.list, 
                       multi_terms_nomin_sent2.list))
nomin_presid.summary=tbl_df(nomin_sent.list)%>%
  group_by(File)%>%
  summarise(
    anger=mean(anger),
    anticipation=mean(anticipation),
    disgust=mean(disgust),
    fear=mean(fear),
    joy=mean(joy),
    sadness=mean(sadness),
    surprise=mean(surprise),
    trust=mean(trust),
    negative=mean(negative),
    positive=mean(positive),
    term_type=unique(term_type)
  )
nomin_presid.summary=as.data.frame(nomin_presid.summary)
rownames(nomin_presid.summary)=as.character(nomin_presid.summary[,1])
km.res=kmeans(nomin_presid.summary[,-c(1,ncol(nomin_presid.summary))], iter.max=200, centers=2)
fviz_cluster(km.res, stand=FALSE, repel= TRUE,
             data = nomin_presid.summary[,-c(1,ncol(nomin_presid.summary))], 
             xlab='', xaxt='n', ylab='', show.clust.cent=FALSE)

table(nomin_presid.summary$term_type,as.vector(km.res$cluster))
##         
##          1 2
##   multi  6 4
##   single 6 4

From the outcomes of clustering shown above, the k-means algorithm cannot successfully classify those Presidents into two groups by just using the mean values of each sentiment.

Conclusion:

From above analysis we can conclude that there is not much but still a little distinct difference between one-term Prsidents and multi-terms Presidents in their nomination and inaugural speeches, especially in the aspects of word frequency, length of sentence and sentiments.

From the word frequency analysis, the most popular words are roughly the same in both inaugural and nomination speeches. However, thers are still some words, like ‘american’ and ‘new’, appeared much ofter in nomination speeches with those Presidents who performed multi-terms. Although this is subtle difference which can be hardly seen, perhaps it changes what voters think.

From the analysis of length of sentences, we find that almost all the Presidents prefered short sentences within 15 words in their speeches. More specifically, Presidents who owned multi-terms enjoyed a slightly longer sentences and more diverse length of sentences in their speeches.

From sentiments analysis, we find something interesting that Presidents who performed only one term more obviously displayed their sentiments in both kinds of speeches than those with multi-terms, especially in the senses of anger, disgust and trust. It is intuitive since performing a speech is a direct way for Presidents to convey their thoughts and ideas towards people, exaggratively indicate their sentiments would conduct into a opposite outcome.